Canon 7D Auto Focus and Imaging

[colm boran]

<p>I borrowed my brother's 7D camera body a few months ago for an outdoor portrait shoot, but my initial plan of shooting in the late afternoon sun was thwarted when heavy clouds rolled in right as I started shooting. From my experience with my 20D, I was nervous about noise if I increased the ISO too high, so I left it at a fairly low setting (400), and took my chances with slowish shutter speeds. I was shooting with the camera mounted to a tripod, and from experience I knew if I kept the adult model still, I could shoot shutter speeds at 1/30 s or so, and be alright.</p>

<p>Although I loved the features and handling of the 7D, I didn't feel the images were better, and perhaps not even as good, as what I would have obtained with my 20D. Everything seemed a bit soft and the focus seemed to be off in several images. I was kicking myself for breaking my own rule about never using new equipment for the first time on an important shoot. But, I chalked up the soft images to the slow shutter speeds and possible model movement, and dismissed the whole thing as operator error.</p>

<p>Weeks later, a friend of mine loaned me his 7D after complaining about less-than-spectacular images he got on his Hawaiian vacation. Coincidentally, I had just received a brand new EF 85mm f/1.8 from B&H that day, so I mounted that lens on his camera to test both at the same time. I had read reviews that this lens may be a little soft when shot wide open, so I took some tripod-mounted photos of the bar code on the lens box. The photos at f/1.8 were indeed pretty soft, but as I stopped down toward f/4, they became very sharp. The troubling part, was it looked like the f/1.8 shots were out of focus, not just soft due to the optics. Out of curiosity, I switched to Live View with 10x magnification, and manually focused. The f/1.8 shots were now very sharp! After going through the range auto focus of microadjust settings, I found that the sharpest focus was achieved at a setting of +15 with the 85mm f/1.8 lens. I began to wonder if my portrait shoot using my brother's 7D might have gone better had I calibrated my lens beforehand ...</p>

<p>Fast forward to today, and I am considering buying a friend's EF 70-200mm f/4L IS. He took his 7D back after getting fed up with focusing issues on 3 different 7D bodies that he had tried, and he ordered a Nikon D700 instead. He's now selling his Canon lenses. He loaned me his 70-200 for a few days to try it out to see if would like it. I plan to upgrade my 20D to either a 7D or the 7D's eventual successor, so I borrowed my brother's 7D again. I took some test photos of my cat's vertical cylindrical scratching post (made with sisal rope) using the following lenses:</p>

 

 

<p>300mm f/4L IS</p>

 

 

<p>70-200 f/4L IS</p>

 

<br />

 

<p>I focused each shot on the center of the scratching post. The vertical cylinder is a good target because you can see if the lens is front- or back-focused by seeing which part of the cylinder is in focus. If the sides of the cylinder are more in focus than the front, the lens back-focused. If nothing is in focus, you can assume it front-focused.</p>

 

 

<p>All images:</p>

 

<ol>

<li>were shot with the lens at maximum aperture (wide open) to minimize depth-of-field and better determine how well the camera was focusing on the target </li>

<li>using Auto Focus were shot with the single, center Auto Focus point selected</li>

<li>of the scratching post were made with the camera or lens mounted on a tripod, with the mirror locked up and the 2-second timer active to minimize camera shake and mirror-induced vibration</li>

<li>shot on the tripod had Image Stabilization (IS) disabled</li>

<li>shot hand-held had IS enabled (Mode 1)</li>

<li>were routed through my usual workflow: raw capture, processed in Adobe Lightroom 3.3 with Adobe Camera Raw conversion (sharpening set to narrow (faces))</li>

<li>were cropped at 100% magnification to show the pixel-level image</li>

</ol>

<strong>

<p>Auto Focus Microadjustment</p>

</strong>

 

<p>I varied the Custom Function Auto Focus Microadjustment parameters from -20 to +20 in increments of 5. Between each auto-focused image, I manually de-focused the lens, alternating between back-focused and front-focused.</p>

 

 

<p>Initial Conclusions:</p>

 

<ol>

<li>The images made with the telephoto L lenses didn't really benefit from auto focus microadjustment (i.e., the best images were with settings very close to the default (zero)).</li>

<li>There is focus error variability from frame to frame, even when the focus microadjustment wasn't altered between frames. The differences could be repeatability error in the auto focus mechanism, but in some cases there seemed to be more consistent error when the focusing was recovering from front-focused points rather than back-focused points.</li>

<li>The 70-200 f/4L lens appeared sharper than the 300mm f/4L when shot at the optimum microadjust settings.</li>

</ol>

 

<p><strong>Auto Focus Repeatability</strong></p>

 

 

<p>To further test focus error repeatability, I set the microadjust setting at 0, and I shot 5 auto focus frames one after the other, always manually back focusing between each shot. Then I shot 5 more auto-focus frames, always manually front focusing between each shot.</p>

 

 

<p>I was curious if the 7D had more focus error repeatability than my old 20D, so I put the 70-200 zoom on the 20D and shot 5 auto-focused frames with manual back-focusing between each frame, and 5 more with manual front-focusing between each frame.</p>

 

 

<p>Initial Conclusions:</p>

 

<ol>

<li>The 7D has auto-focus error variability (i.e., the camera's auto focus is noticeably unable to repeat the focus from one shot to the next), but nearly all the shots were focused well enough to be considered acceptable in my opinion</li>

<li>The 20D auto-focus error variability is similar to the 7D.</li>

<li>The 20D images appear generally as sharp as the 7D.</li>

</ol>

<strong>

<p>Hand-Held Shooting</p>

</strong>

 

<p>Since I shoot a fair amount of sports and candid portraits, I can't always shoot on a tripod, with the mirror locked up, and on self-timer. So I also shot the bird feeder in the back yard, hand-held, with the same lenses, plus a couple more non-L lenses that I own (EF-S 17-85mm f/4-5.6, and EF 85mm f/1.8). All images were made at maximum magnification (zoomed in), with wide open aperture. I shot with both the 7D and the 20D with all four lenses.</p>

 

 

<p>Initial Conclusions:</p>

 

<ol>

<li>There is no significant difference in the focusing error between the 20D and the 7D in my handheld quick shots.</li>

<li>The 20D images appear generally as sharp as the 7D.</li>

</ol>

 

<p>You can see the complete results for yourself here: <a href="http://www.number6photo.com/CanonAutoFocus/index.html">http://www.number6photo.com/CanonAutoFocus/index.html</a></p>

 

 

<p>Titles and captions for each image show the equipment and settings.</p>

 

 

<p>I'm pretty sure the somewhat soft images issues I had during my initial outdoor portrait shoot were likely due to not calibrating my non-L lens that I was using, combined with the slow shutter speeds, some camera and model movement. Perhaps more importantly however, looking at the 7D's 18 megapixel 100% crops, and comparing them to the 20D's 8 megapixel 100% crops is a little unfair. Camera movement and focus issues are amplified at the pixel level on the 7D compared to the 20D. I'm happy to report that despite looking a little soft at 100%, normal-sized prints of my outdoor shoot made from the 7D with typical crops looked just fine. It does make you wonder why Canon bothered making the 7D's imager 18 megapixels vs 8, 10 or 12 though. Seems like a waste of memory, and the processing engine has to work that much harder to shove all those bits through the pipeline. Plus, larger photosites from fewer pixels on an APS-C-sized sensor, combined with Canon's noise-processing prowess, should lead to even better noise performance.</p>

 

 

<p>I have seen some issues posted by others where the focus on their 7D bodies appeared much worse than the two 7D bodies I have tried, and I have no doubt that my buddy that turned his 7D in for a Nikon D700 had genuine problems. I'm not sure I'm ready to pull the trigger on purchasing my own 7D yet. I'm interested in the improved AF tracking and frame-per-second capabilities of the 7D compared to my 20D for the occasional sport shoots that I do, but the images I have produced from the two 7D bodies I have borrowed have not been significantly better than my 20D, so it's hard to justify the upgrade. I think I may like the 5D Mark II, but I don't think the AF has fast enough sport-shooting capability (specs seem similar to my 20D). If I didn't already own so much Canon glass, I would consider get the Nikon D700, which seems to have the auto focus and higher frame-per-second shooting, combined with a full frame sensor at a sensible 12 megapixels. I would love to compare the auto focus capability of the Nikon 700D vs. the Canon 7D. Maybe I'll have a chance after my friend's 700D arrives.</p>

 

[philip_wilson]

<p>I am surprised at your issues with the 7D as I find mine works well. I find that the AF on my 7D works as well (except in very low light) as my 1 DIIN's AF. Indeed the picture quality of the 7D beats the 1DIIN except at ISO 1600 and 3200 where the images are similar (the 1DIIN is less noisy but the extra pixels of the 7D allow you to achieve similar IQ). In general I find that the 7D is as good a sports shooter as the 1DIIN - except in terms of body construction. If you want to shoot sports and considered the 1DIIN to be better than the 20D (Canon certainly positioned it that way) then the 7D will also prove to be better.<br>

My 5DII does produce clearly better images than the 7D - even at low ISO but the AF is not great for sports. It is very usable but you will get more out of focus shots with the 5DII than the 7D or a 1 series camera.<br>

It is possible you have a bad body but I find that most supposed AF issues come down to user error - either using AI Focus or having lots of AF points active. In your case they do not seem like likely explanations.<br>

As to the lenses, I own (among others) the 70-200 F4 L IS, 85 F1.8 and 300 F4 L IS. Of the three the 85mm is the sharpest and the other two lenses are very similar ( maybe the zoom is slightly sharper in the centre).<br>

DXO mark rates the 20D at 41 lp/mm with the 85 F1.8, the 7D at 47 and the 5DII at 66. Similarly with the 70-200 F4 LIS they give the 20D 42 lp/mm, the 7D 48 and the 5DII 65.<br>

As to the sensors DXO Mark gives the 20D and 7D similar signal to noise performance but the 14 bit 7D has a better dynamic ranger up to ISO 800. Of course the 7D has lots of other advantages over the 20D beyond frame rate. It has a better menu system, live view, a high resolution 3 inch screen and of course video. Here are two high ISO crops of Hockey shorts taken during a game - one with the 1DIIN, the other with the 7D. They both represent the same area of the image but the 7Ds will obviously look bigger as it has more pixels. Printed at the same size the 7D is probably the sharper image.</p><div></div>

[philip_wilson]

<p>7D </p><div></div>

[philip_wilson]

<p>Finally just to show how good the 5DII is here is the 5DII equivalent</p><div></div>

[philip_wilson]

<p>If all these crops are the same size then you can see the 5DII is a clear winner but the extra resolution of the 7D results in a slightly sharper image than the 1DIIN when they images are the same size. All of these shots were taken with the 70-200 F2.8 handheld.</p>

[lester_wareham]

Indeed I went through a similar loop with the 7D, I too discovered the focus variability I had been aware of on the 20D

was similer on the 7D - this just seems to be part of using AF camera, even so I am sure they do a better job than I

achieved with my manual focus FD cameras.

 

I too used the micro adjust on the 7D, interestingly one lens that had been lacklustre onthe 20D is now excellent on

the 7D, it is clear to me I was generally lucky with the combination of tollerances on the 20D and my lens set and less

so with the 7D, again this just reflects the mass production tollerencing.

 

Re the 18mp sensor, this does provide more reach with very sharp telephotos combined with good technique, if you

don't need the resolution the you could use mraw for example.

[PuppyDigs]

<p>I've used a 7D extensively for 15 months and haven't experienced problems with AF. In fact, once I learned to "work" the 5 AF modes (took many weeks!), I've found AF to be the best of any EOS I've owned. Initially I suffered some focus misses in auto select, single point, 3-zone and cluster when shooting close stationary objects. Basically a single AF point or small group is so large it can lock on either an eye or nose (or center or side of a curved object), resulting in seemingly random focus errors. Then I discovered spot focus and all these problems disappeared. The greatly reduced AF point size propels my macro AF into a near 100% hit rate. Obviously spot AF isn't appropriate for all subjects--I rarely use it for landscape or wide angle--but I mapped spot focus to the DOF button so I can engage it on the fly in situations I know may trick normal AF. I probably use spot AF about 1/3 of the time.</p>

<p>And, this really isn't a 7D problem. I suffered similar misses due to large AF points with my 10D, 20D, 40D, 50D and 5D. The 7D is the first camera I've owned that I could reduce the AF point size and defeat AF twitching from eyebrow to eye from shot to shot.</p>

<p>Oddly, none of my current lenses need MA tweaks.</p>

[colm boran]

<p>@Puppy Face: Intrigued by your results with Spot AF, I mounted my 300 f/4 L again, and shot the scratching post cylinder again with Spot AF, and Single Point AF; 8 shots each, alternating between manual front and back de-focusing between each shot. Oddly enough, all the shots came out essentially the same this time; I can't discern any significant difference between any of the successive frames, much less whether they were shot with Spot AF or Single Point AF.</p>

<p>You can see the results here: <a href="http://www.number6photo.com/CanonSpotAFversusSingle-PointAF/index.html">http://www.number6photo.com/CanonSpotAFversusSingle-PointAF/index.html</a></p>

<p>I don't know why it did this well, although I did notice that I had inadvertently left the Image Stabilization ON despite being mounted on the tripod. I had heard on older lenses like my 300 f/4L IS that leaving IS on while holding the camera absolutely still (e.g., when mounted on a tripod) could potentially degrade the image, but these images are as sharp, if not sharper, than the previous attempts on the tripod with IS OFF. 2s timer and mirror lock up were used in all cases, so I'm at a loss to explain it.</p>

<p>I also did some quick tests using Spot AF with my new EF 85mm f/1.8 using the MA focus custom function. On my brother's camera, the lens does very well anywhere between -5 and +5, so I plan to leave it at 0. I'll see if I can borrow my friend's 7D again and use Spot AF on his to see if it makes a difference.</p>

<p>If the 7D can focus this well all the time, I would be thrilled. Thanks for the tip!</p>

[dcstep]

<p>I have around 25000 clicks on my 7D with no problems.<br>

No one mentioned Single-Shot vs. Continual Tracking. I generally leave mine on tracking, but when doing portrait or archetectural I'll turn it to single shot. If you're using single-point AF and you put the point somewhere on the subject that has little contrast it can hunt a bit. For a face, I'll use the 19-point and make sure that an active point is on the eye. You can use single-point manual selection to make certain that you're on the eye. If you get in really close you'll need single-point manual selection instead of the 19-point which needs a bit of distance.</p>

<p>For birds I found that I needed different AI servo speed selections with my 500mm f/4L depending on whether I was using my 1.4x TC or not. The key to finding the best selections for me was to experiment with the parameters. The 7D's AF modes are very flexible, but you need to work with them to understand what the camera is going to do.</p>

[robertjewett]

<p>https://www.lensrentals.com/news/2008.12.22/this-lens-is-soft-and-other-myths</p>

[colm boran]

<p>@Robert: That's a good article, and no doubt many will be helped by it. It certainly applies to my experience with my 85mm f/1.8 on my friend's 7D. But it doesn't address the repeatability issue that I found on my brother's 7D. Without changing any settings, or camera position, etc., the focus error changes enough to be noticeable (at the pixel level) from one frame to the next. Or at least it did one day, and not the next, which is a little perplexing.</p>

<p>I think it's this inconsistency that shakes people's confidence. If you the camera / lens combo always does one thing, you can use the micro adjust feature to fix it for good. But if it's front focusing one shot, and back focusing the next shot, and it's difficult to see if it did either one on the LCD display, you won't know if you've got or missed the shot until you get home.</p>

<p>For moving targets like sporting events, no one expects 100% of the images to be perfectly focused. Most of us have to admit that the camera is doing a much better job than we would do if we were focusing manually, especially at 8 fps. But for static images like portraits or landscapes, most of us were probably able to get 95% or more shots sharply focused manually, and we expect the camera that can do such an incredible job on moving targets to do a damn-near perfect job on static subjects. In my experience, neither my old 20D or the two new 7D's I have tried are consistently nailing the static shots.</p>

<p> I'm beginning to think the auto focus errors have always been there. For example, my simple comparison showed that my aging 20D, which has been consistently very good with static subjects over the years, is in reality no more repeatable than the 7D, which I initially thought was erratic. I think it's simply because the auto focus errors are more visible on modern cameras (at the pixel level), especially 1.6x crop bodies with 18+ megapixels. Perhaps that is also why people shooting 20+ megapixel full-frame bodies don't notice it as much. </p>

<p>As I mentioned, the focus error in reality is very small, and not noticeable in the final image (assuming only moderate cropping was applied to the original image). I do believe that 18 megapixels on a 1.6x crop body has reached the point of diminishing returns however; whether from the optical path not being able to resolve any more detail, or whether it's from auto focus error repeatability, or a combination of the two.</p>

[dcstep]

<p>Colm, are you shooting in single-shot mode when you experience these shot-to-shot differences?</p>

[colm boran]

<p>David, yes, I was shooting in single-shot mode, with manual de-focusing between each shot. I did some earlier comparisons with AI Servo and saw no difference. For portrait shooting, I think single-shot mode is the only realistic alternative if you want to focus (e.g., with the center AF point), compose, and then release the shutter.</p>